Grove: Distributed ML Training Over Apple AirDrop — No Central Server Required

Swarnim Jain / Hacker News·low signal

A new Show HN project demonstrates distributed machine learning training using Apple's AirDrop protocol to coordinate across devices, eliminating the need for a central server. At 32 points on HN, the approach leverages existing Apple Silicon hardware and MLX for privacy-preserving collaborative training. Early-stage but signals growing interest in decentralized ML training as an alternative to cloud-based approaches.